To enhance the long-term displacement prediction accuracy of landslides, the GCformer model was applied to landslide displacement forecasting, and a novel landslide displacement prediction approach grounded in the GCformer model was proposed. This methodology leveraged rainfall and displacement as input variables, utilized the GConv_msk module to capture the global information of the sequence, and combined a linear scaling technique of sequence length to efficiently extract data features. Concurrently, the PatchTST model was employed to automatically extract short-term and long-term signals from the sequence data, in order to obtain more comprehensive historical information and bolster the model's robustness and modeling capability. Finally, the landslide displacement monitoring data from Jinliuping Village and Yuanshitan Village in Huichuan County, Dingxi City, Gansu Province, were utilized for case validation. The findings demonstrate that the proposed model exhibits superior prediction accuracy and reliability. In comparison to the Autoformer model and the FEDformer model, the GCformer model is found to achieve the lowest error in both total displacement and vertical displacement.